This is a good opinion, on two different fronts.
First, the majority opinion (written by Judge Thompson) and the dissent (written by Judge Betty Fletcher) differ substantially regarding their confidence in eyewitness testimony. The majority holds that any Brady error would be harmless in light of the eyewitness identification of the defendant by several witnesses. This was essentially the only evidence at all against the defendant -- who was convicted of bank robbery -- but the majority is convinced that the defendant was clearly the perpetrator based upon this testimony. By contrast, Judge Fletcher is much less convinced of the accuracy of such testimony, particularly in light of the absence of any supporting evidence and the fact that someone else -- who looked a lot like the defendant -- had been convicted of nearby bank robberies, raising the distinct possibility of misidentification.
Judge Thompson and Judge Fletcher both do a good job of supporting their positions. It's a very, very good case in which to assess -- in a concrete fashion -- how much one believes in eyewitness testimony.
Second, there's also a more subtle difference in judicial approach. For Judge Fletcher, the possibility of an innocent person being incarcerated is an extremely big deal. So her approach is to say: "Grant a retrial and let the jury hear the evidence about the alternative perpetrator. If they convict, fine. At least we'll know that the jury had all the evidence. By contrast, the risk of having convicted an innocent person here is too great." By contrast, for Judge Thompson, efficiency concerns are paramount. His approach basically says: "Look, no trial is perfect, and I'm pretty convinced the defendant is guilty. No need to spend the time and money on a new trial that I think will come out the same way (or, worse still, if witnesses may have died or become stale, might result in an erroneous acquittal)."
Two very competing approaches. A nice, relatively short, and informative and significant set of opinions. Definitely worth reading.